Expression of Noncovalent Hepatitis C Virus Envelope E1-E2 Complexes Is Not Required for the Induction of Antibodies with Neutralizing Properties following DNA Immunization

Abstract

Interactive glycoproteins present on the surface of viral particles represent the main target of neutralizing antibodies. The ability of DNA vaccination to induce antibodies directed at such structures was investigated by using eight different expression plasmids engineered either to favor or to prevent interaction between the hepatitis C virus (HCV) envelope glycoproteins E1 and E2. Independently of the injection route (intramuscular or intraepidermal), plasmids expressing antigens capable of forming heterodimers presumed to be the prebudding form of the HCV envelope protein complex failed to induce any significant, stable antibodies following injection in mice. In sharp contrast, high titers of antibodies directed at both conformational and linear determinants were induced by using plasmids expressing severely truncated antigens that have lost the ability to form native complexes. In addition, only a truncated form of E2 induced antibodies reacting against the hypervariable region 1 of E2 (specifically with the C-terminal part of it) known to contain a neutralization site. When injected intraepidermally into small primates, the truncated E2-encoding plasmid induced antibodies able to neutralize in vitro the binding of a purified E2 protein onto susceptible cells. Because such antibodies have been associated with viral clearance in both humans and chimpanzees, these findings may have important implications for the development of protective immunity against HCV.

Hepatitis C virus (HCV) is the major causative agent of transfusion-associated and community-acquired non-A, non-B hepatitis worldwide (6, 22). More than 70% of HCV infections become chronic, with a significant risk in 5 to 20% of cases of progression to liver cirrhosis (1) and hepatocellular carcinoma (33). Only 20 to 30% of long-term responses occur in patients treated with alpha interferon (IFN-α), the currently used therapy (15). The development of new therapeutic agents as well as a vaccine for prevention or treatment of HCV infections has become a priority. A first step in designing a vaccine is the identification of both host and viral components involved in the development of neutralizing immunity. In the HCV model, such protection may in part be due to neutralizing antibodies targeted at the envelope glycoproteins E1 and E2. Successful in vivo protection of chimpanzees has been achieved following immunization with recombinant E1 and E2 proteins and has been linked to the induction of specific anti-E2 antibodies (5). Such antibodies neutralizing in vitro the binding of purified E2 onto susceptible cells, referred as “neutralizing of binding” (NOB) antibodies (32), have recently been linked to the resolution of chronic infection in humans (21). Several observations have shown that the hypervariable region 1 (HVR-1) of E2 contains an important neutralization domain. In particular, antibodies present in the sera of infected patients or induced by immunization and targeted at this region can prevent viral infection in cell cultures (37, 44). In contrast to anti-E2 antibodies, to date, the participation of anti-E1 antibodies in viral clearance remains undocumented.

Various studies using transient viral and nonviral expression systems have shown that HCV envelope glycoproteins E1 and E2 interact to form complexes (17, 29). Two forms of E1-E2 complexes are detected: heterogeneous disulfide-linked aggregates formed by misfolded proteins and heterodimers stabilized by noncovalent interactions composed of native glycoproteins (8, 10). The latter have been proposed as the prebudding form of the HCV envelope glycoprotein complex. Conformation-sensitive E2-reactive monoclonal antibodies (MAbs [H2 and HMAb 503]) have recently been described which selectively recognize noncovalently associated complexes, allowing the distinction to be made between native complexes and misfolded aggregates (8, 18). As described for human immunodeficiency virus envelope proteins (11, 31), interactions between HCV glycoproteins could affect epitope presentation and have an important influence not only on the antigenicity of the proteins but also on their immunogenicity.

Genetic immunization, which allows the de novo synthesis of the DNA-expressed antigens in the host’s cells (42), has been shown to elicit both protective humoral and cellular immune responses in several animal models of viral infection (2, 30, 39, 40). This vaccination mode, similar to strategies based on the use of attenuated viruses or live expressing vectors, provides the biological context for antigens to be naturally processed with respect to posttranslational modifications, protein folding, and assembly (38). The opportunity for de novo-synthesized proteins to achieve proper maturation is a particularly important element in the case of proteins that require the help of additional partners to fully mature. An example of such proteins are proteins constituting viral envelopes. These proteins, usually glycoproteins, often display complex interactions between themselves and/or cellular partners for the constitution of functional, native envelope complexes (16, 19). The interactions between HCV E1 and E2 proteins thus offer a good model to study the advantages and limitations of DNA-based immunizations for the induction of antibodies directed at antigenic structures existing as complexes and representing critical components of a vaccine (5, 21).

Here, we report on the efficacy of different plasmids engineered to favor or limit the formation of E1-E2 complexes at inducing specific antibodies and cytokine release. We showed that expression of presumed native E1-E2 complexes failed to induce any significant humoral responses, whereas optimal responses (including anti-E2 antibodies with neutralizing of binding activity) were obtained in mice and primates with truncated forms of the proteins.

MATERIALS AND METHODS

Plasmids and in vitro expression studies.

E1 and E2 sequences were amplified from a vector containing the full-length cDNA sequence of the HCV-H strain, 1a and cloned into the XbaI-NotI or SmaI sites of the pCI vector (Promega), resulting in the pCI-based vectors (Fig. 1). Dicistronic expression vectors, allowing the coexpression of two distinct genes from two transcripts within the same cell, were generated with the vector pFX (generous gift from M. Nasoff), which contains two cytomegalovirus (CMV) promoters followed by one PacI or NotI site for the cloning (Fig. 1). All plasmid-cloned fragments were verified by sequencing (34).

FIG. 1.

Plasmids and in vitro expression studies. (A) Schematic representation of the HCV envelope region. Black boxes correspond to signal peptide sequences. Sequences expressed by the indicated plasmids are shown diagrammatically by bars. Empty boxes represent the two CMV promoters of the pFX vector. (B) Quantitative determination of intracellular expression of E1 and E2 antigens by capture ELISA. CAT determination was realized after cotransfection of pcDNA3/CAT with each plasmid to standardize transfection efficiency. Results are given as OD values. (C) Identification of HCV envelope proteins and complexes. Transfected Cos-7 cells were labeled with [³⁵S]methionine, and cell lysates were immunoprecipitated with either the anti-E2 H2 MAb or the 503 HMAb, and representative results are shown. Transfected plasmids are indicated at the top of each lane, T−, pCI. Samples were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (13% polyacrylamide). The positions of the ¹⁴C-labelled protein markers are shown on the left, and those of HCV-specific proteins are shown on the right.

Expression of E1 and E2 antigens was examined in cell extracts and supernatants by using a capture enzyme-linked immunosorbent assay (ELISA) employing highly specific monoclonal antibodies (23a) as well as by immunoprecipitation following transient transfection of Cos-7 cells. For the capture ELISA, cells from three 35-mm wells were pooled and lysed in a mixture of 10 mM MOPS (morpholine propanesulfonic acid [pH 6.5]), 10 mM NaCl, 1 mM EGTA, and 1% Triton X-100. Cell lysates were then used to coat microtiter plates at a ½ dilution. After blocking, MAbs directed against either E1 or E2 were incubated. The MAb was detected with a goat anti-mouse peroxidase conjugate. Cotransfection of the chloramphenicol acetyltransferase (CAT)-expressing plasmid pcDNA3/CAT (Invitrogen) was employed to monitor transfection efficiency. Transfected cells (Lipofectamine plus; Gibco BRL) were metabolically labeled at 24 h posttransfection with 200 μCi of [³⁵S]methionine per ml during 24 h and isolated by using the MACSelect-transfected cell selection kit (Miltenyl Biotec). Cell suspensions were then centrifuged, and the pellets were lysed with 0.5% Nonidet P-40 in a mixture of 10 mM Tris-HCl (pH 7.5), 150 mM NaCl, 2 mM EDTA, 20 mM iodoacetamide, and 10 μg of aprotinin per ml. HCV-specific proteins were immunoprecipitated from the clarified lysates as described previously (9) by using two anti-E2 MAbs that selectively recognize E1-E2 noncovalent complexes: a murine antibody, H2 (8), and a human one, 503 (18). Immunoprecipitates were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis.

DNA-based immunization. (i) Mice.

Six- to eight-week-old female BALB/c mice (n = 5 or 6 per group) were purchased from Charles River. All DNA preparations were produced with endotoxin-free purification columns (Qiagen). Immunizations were performed with either a gene gun (PowderJect), resulting in the injection of 5 μg of plasmid DNA into the abdominal skin (referred to as intraepidermal [i.e.] injection) or with a syringe in the anterior tibialis muscle (referred to as intramuscular [i.m.] injection) by using 100 μg of plasmid DNA per injection at weeks 0, 9, and 21 as previously described (13, 24).

(ii) Tamarins.

Six adult female tamarins (Saguinus labiatus) seronegative for HCV were individually housed and cared for according to approved standard operating procedures. Animals were tranquilized for inoculations and blood collections. Four animals were injected with 20 μg of total DNA per injection per boost in the abdomen with the gene gun, and the two other animals were immunized intramuscularly in the tibialis muscle of one leg with 400 μg of DNA in saline at weeks 0, 5, 9, and 20.

Antibody titers and isotypes.

Induced antibodies were measured with a specific ELISA (INNO-test for anti-E1 and anti-E2 antibodies; Innogenetics) as previously described (23a, 28). Sera from tamarins were analyzed with an antihuman immunoglobulin G (IgG) Fe labeled with horseradish peroxidase (Dako) used as a secondary antibody. Antibody titers were calculated as the serial threefold dilution which gave an optical density (OD) that equaled the cutoff. For mice, the cutoff value was established as the mean OD + 3SD of 10 sera obtained from control mice, while for tamarins, it was determined for each animal as the equivalent to 3× the OD of sera obtained before the primary injection.

Peptide-based epitope mapping.

Peptide A1H encompassing the HCV-H hypervariable region (HVR-1) of the E2 protein, located between amino acids (aa) 384 and 411 on the polyprotein (41), was used in the ELISA as previously described (13). Sera were tested at a 1:100 dilution, and only those giving absorbance values >2.5 times the absorbance value of negative sera were considered positive. Nineteen decamers overlapping by 9 aa and encompassing the A1H peptide were used: N1 to N8 for N-terminally-located peptides and C9 to C19 for C-terminally-located peptides. ELISA plates were first coated with streptavidin (Sigma), and assays were performed as described above by using the 19 peptides, each of which carried a biotinylated spacer peptide at its N terminus (Neosystem).

Neutralization assay.

Mice and tamarin sera were tested for their ability to neutralize the binding of the E2 protein onto MOLT-4 cells in the recently developed NOB assay (32). Quantification of NOB antibodies was performed as previously described (21), by incubation of a suboptimal concentration of biotinylated recombinant CHO cell E2 protein (1 μg/ml) with different dilutions of the tested sera. E2 binding to target cells was detected with a streptavidin-phycoerythrin conjugate (2.5 μg/ml).

Cytokine measurements.

In vitro cytokine production was analyzed at 32 weeks post primary injection following antigen-specific in vitro stimulation of splenic cells from immunized mice as previously described (13). Cells were stimulated with the same recombinant E1 and E2 proteins used for antibody titration, added at final concentrations of 2 or 5 μg/ml. Culture supernatants were harvested at 24 h for interleukin-2 (IL-2) and 48 h for IL-4, IL-5 and IFN-γ testing, and cytokine levels were measured by quantitative ELISA (Biotrak; Amersham). All cultures were performed in duplicate, and the data indicated represent the ranges of cytokine concentration obtained with the different concentrations of antigens.

RESULTS

In vitro expression studies: different plasmids can favor or limit the formation of complexed structures.

In order to favor or, on the contrary, to prevent E1-E2 interactions, we designed several plasmids expressing either the full-length envelope region of HCV or full-length and/or truncated forms of each envelope protein. These included monocistronic vectors expressing the full-length envelope region (pCI E1E2p7), E1 or E2p7 proteins (pCI E1 and pCI E2p7), or truncated forms of the proteins lacking their transmembrane domains (pCI E1t, pCI E2t, and pCI E1E2t) as well as their counterpart dicistronic plasmids (pFX E1-E2p7 and pFX E1t-E2t) (Fig. 1A).

Expression of E1 and E2 was confirmed by immunofluorescence analyses after transient transfection of Cos-7 cells. Diffuse cytoplasmic localization of both antigens was observed independent of the plasmid used (data not shown). Quantitative determination of the intracellular expression of E1 and E2 showed differences between the amounts of antigen produced by the different plasmids (Fig. 1B). Expression of the truncated form of E1 led to higher antigen production than that obtained with the full-length form of the antigen. Differences in E2 production between the plasmids used were less pronounced with the exception of one plasmid, pCI E1E2t, which led to a very low production of E2 protein. Antigens were not detected in the supernatants of transfected cells (i.e., secreted antigens), even when the E1 and E2 proteins were expressed without their hydrophobic anchor domains (from aa 311 for E1 and aa 675 for E2). This suggests that the secretion level of the expressed antigens, if secretion does take place, is very low or that the proteins are unstable. Overall, under the experimental conditions used, the capacity to produce secreted antigens does not appear dramatically different between the plasmids, the E1 and E2 proteins remaining mainly intracellular.

Immunoprecipitation studies revealed that when expressed as full-length proteins either from a monocistronic plasmid, pCI E1E2p7 (Fig. 1C, lane 2), or from a dicistronic plasmid, pFX E1-E2p7 (data not shown), E1 and E2 have the capacity, as expected, to form noncovalent heterodimers coprecipitated selectively by a conformation-sensitive MAb. No E1-E2 noncovalent complexes were detected any longer when the antigens were expressed from plasmids encoding C-terminally truncated forms of one or both proteins (pCI E1E2t and pFX E1t-E2t), as shown in the figure (lanes 3 and 4). The lack of immunoprecipitation with the pCI E1E2t plasmid is in accordance with previous reports (14, 26, 36) which indicate that the carboxy terminus of E2 is critical for the E1-E2 interaction. The absence of coprecipitation of a truncated form of E1 with a truncated form of E2 indicates that if E1t and E2t form complexes, they are not properly folded, as suggested by Michalak et al. (26), and therefore are not recognized by the conformation-sensitive HMAb 503. Thus, the ability of the different plasmids to express native interactive proteins could be governed by the cloning strategy implemented.

Humoral responses to HCV E1 and E2 antigens in mice: a strict dependence on the form of the DNA-expressed antigens.

For all monocistronic pCI constructs, DNA immunizations were performed by using the intraepidermal route, which has been shown to result in higher antibody titers than the i.m. route (13, 28). For the dicistronic pFX constructs, we compared three injection modes: i.e., i.m., and a combination of both (i.e. + i.m.).

Seroconversion rates and antibody titers are indicated in Fig. 2. Overall, both antibody titers and seroconversion rates were much higher for E2-encoded plasmids than E1-encoded ones. Except for four animals that displayed titers >1:2,000, all others had no detectable or extremely low anti-E1 antibody titers, whatever the E1-encoding plasmid injected. This observation suggests that E1 is a very poor immunogen and limits the evaluation of the induced anti-E1 responses.

FIG. 2.

Antibody (Ab) titers and seroconversion. Each plasmid indicated on Fig. 1 was injected into one group of mice either by the i.e. route (monocistronic constructs) or by one of the three injection modes, i.e., i.m., or a combination of both (i.e. + i.m.). Median titers, given for seroconverted mice only, are represented. Results are shown as the reciprocal of the serum dilution equivalent to 3× the OD of sera from mice injected with the vector pcDNA3. The numbers shown on top of each bar indicate the number of animals that seroconverted per animal group. (A) Responses to proteins expressed by monocistronic pCI-derived constructs. (B) Responses to proteins expressed by dicistronic pFX-derived constructs. p.i., post-primary injection.

Despite a total of three injections and independent of the injection route, some E2-encoding plasmids never produced any detectable antibodies or produced antibodies only at very low titers: pCI E2p7, pCI E1E2p7, and pFX E1-E2p7 (maximal median titers ranged from 1:40 to 1:8,510). These results were obtained with both mono- and dicistronic plasmids and corresponded to plasmids containing the full-length sequence of E2. Mouse sera were in addition used to perform immunoprecipitation studies with cell extracts expressing E1-E2 complexes (obtained by infection with recombinant vaccinia viruses expressing such complexes). No positive signals could be detected when sera corresponding to mice injected with plasmids expressing full-length E1 and/or E2 antigens were used (data not shown), arguing for a lack of specific antibodies in these sera. In sharp contrast, plasmids expressing a truncated form of E2 (pCI E2t and pFX E1t-E2t) yielded much higher titers of long-lasting antibodies, with maximal median titers reaching 1:30,000. Such antibodies have been shown in addition to be capable of immunoprecipitating E1-E2 complexes (28). One notable exception was the plasmid pCI E1E2t, expressing a full-length E1 antigen together with a truncated form of E2 which induces very low anti-E2 antibody titers and very transient anti-E1 antibodies. The use of dicistronic plasmids was overall equivalent to that of the monocistronic ones. Comparison of the three different modes of immunization used for injection of the dicistronic constructs indicated that immunization modes which included an i.e. route always resulted in higher antibody titers (Fig. 2B). The latter observation is in agreement with previous observations made by using monocistronic E2-expressing plasmids (13, 28). Altogether, the nature of the immunization route could not compensate for the very poor performance of the plasmids expressing full-length E1 and E2 antigens.

Epitope mapping: peptide epitopes within the HVR-1 are mainly recognized by antibodies induced following injection of truncated DNA-expressed E2 antigen.

Because of its association with the induction of neutralizing antibodies (12, 32, 37, 43), the reactivity of all sera against the hypervariable region located at the N terminus of E2 (HVR-1) (41) was more extensively analyzed (Fig. 3). Essentially two plasmids induced antibodies recognizing a peptide corresponding to the 27 aa of the HVR-1, both of them encoding a truncated form of E2 (pCI E2t and pFX E1t-E2t). However, remarkably, there was no correlation between the overall antibody titers detected in the ELISA (Fig. 2 and reported in Fig. 3 for each mouse on top of the bars) and the reactivity against the HVR-1 peptide. In the group of mice injected by the i.e. + i.m. route with the pFX E1t-E2t plasmid, one serum sample with a titer of 1:60 (mouse 5) displayed a good reactivity, while another with a higher titer of 1:6,000 (mouse 2) had no detectable activity against A1H. These data show that independent of the anti-E2 antibody titers, plasmids encoding a truncated form of E2 remain the best at inducing production of antibodies targeted at this domain. Whatever the expression context or the injection route, plasmids expressing full-length proteins fail to induce such antibodies.

FIG. 3.

Reactivity of mice sera against the HVR-1. Sera obtained at 25 weeks post-primary injection from each mouse of the different injected groups were tested against the A1H peptide. The different groups of mice are given along the X axis, while the binding results obtained are given as OD values on the Y axis. The cutoff value is indicated by the dotted line.

When all 46 mouse serum samples were further analyzed for their reactivity against minimal determinant domains mapping within the HVR-1 (using 19 overlapping peptides termed N1 to N8 and C9 to C19 according to their N- or C-terminal location within the HVR-1), it was observed that in addition to plasmids encoding a truncated form of E2, plasmids encoding full-length antigen displayed a low but detectable reactivity against some peptides (i.e., the pCI E1E2p7 and the pCI E1E2t plasmids reacting with peptide C15) (Table 1). Overall, immune recognition was mainly targeted at C-terminally-located peptides independent of the plasmid used. Four of the five reactive peptides (N2, C9, C12, C15, and C17) were located in this region.

TABLE 1.

Epitope mapping of HVR-1

Injected plasmid	Route of injection	Mouse	Antibody binding to peptidea
			N2	C9	C12	C15	C17
E1E2p7	i.e.	1b	ND	ND	ND	ND	ND
		2	−	−	−	−	−
		3b	ND	ND	ND	ND	ND
		4	−	−	−	−	−
		5	−	−	−	−	+
		6	−	−	−	−	−
E2t	i.e.	1	−	−	−	−	−
		2	−	−	−	+	−
		3	+	−	−	−	−
		4	−	−	−	−	−
		5	−	−	−	−	−
		6	+	−	−	−	−
E1-E2p7	i.e.	1	−	−	−	+	−
		2	−	−	−	+	−
		3	+	+	−	+	−
		4	−	−	−	+	−
		5	−	+	−	−	+
	i.m.	1	−	−	+	−	−
		2	−	−	−	−	−
		3	−	−	−	−	−
		4	−	−	−	+	−
		5	−	−	+	+	−
	i.e. + i.m.	1	−	−	+	+	−
		2	−	−	−	−	−
		3	−	−	−	−	−
		4	−	−	−	−	−
		5	−	−	+	−	−
E1t-E2t	i.e.	1	−	+	+	−	−
		2	−	−	+	−	−
		3	−	−	+	−	−
		4	−	−	+	−	+
		5	−	−	−	−	−
	i.m.	1	−	−	−	−	+
		2	−	−	−	−	−
		3	−	−	+	+	−
		4	−	−	+	−	−
		5	−	−	−	−	−
	i.e. + i.m.	1	−	−	−	−	−
		2	−	−	+	−	−
		3	−	−	−	−	−
		4	−	−	+	−	−
		5	−	−	−	−	−

^a+, OD values considered positive; −, negative values; ND, not determined. 

^bMouse dead. 

IgG isotypes and in vitro splenic cytokine production: the form of the DNA-expressed antigens has no influence.

Analysis of anti-E1 or anti-E2 antibodies (at 21 and 31 weeks post-primary injection) revealed marked differences between the isotypes of E1- and E2-specific antibodies, which were independent of the injection route and plasmids used. IgG2a or -2b was detected in E2-reactive antibodies, while IgG1 or IgG3 was observed in anti-E1 antibodies (data not shown).

Inasmuch as the IgG isotype profiles of anti-HCV E1 and E2 antibodies suggested the priming of different T helper cell subsets, direct measurements of Th1-specific cytokines (IFN-γ or IL-2) and Th2-specific cytokines (IL-4 or IL-5) were performed. While insignificant levels of IL-4 and IL-5 were detected, splenocytes from immunized mice showed production of IFN-γ (ranging from 30 to 132 pg/ml for E1 and 34 to 855 pg/ml for E2) and IL-2 (ranging from 27 to 73 pg/ml for E1 and 32 to 162 pg/ml for E2) following antigen-specific in vitro stimulation. The different levels of cytokines produced could not be correlated to a specific plasmid or an injection route. Overall, the form of the expressed E1 and E2 did not affect the isotype of the induced antibodies or the in vitro splenic cytokine profiles.

Humoral immune responses to HCV E2 in tamarins: expression of a truncated form of E2 can induce antibodies with NOB capacity.

To evaluate some of the observations made with the murine model, the pCI E2t plasmid was used to immunize small primates. We immunized six tamarins, four by an i.e. injection route (via the gene gun) and two others by the i.m. injection route.

All four of the tamarins immunized intraepidermally developed an E2-specific antibody response, detected either after the first boost in two animals or following the third boost in the two others (Fig. 4). Antibody titers reached 1:6,000 in two animals (tamarins 1 and 4). The E2-specific antibody titers declined with time to reach low but detectable levels after 3 months of follow-up. Upon i.m. immunization, only one tamarin of the two injected (tamarin 5) developed antibodies early after the primary injection (3 weeks), but this response was relatively short lived and could not be enhanced by additional booster injections. Anti-E2 antibody isotypes were of the IgG2 type, in agreement with those previously observed in mice (this study and references 13 and 28). Only one animal (tamarin 4) displayed reactivity against the HVR-1 peptide, and the contribution of small peptides to the overall responses was basically null in tamarins (no antibodies against any of the 19 decamers were detected).

FIG. 4.

Profiles of anti-E2 antibodies (Ab) detected in tamarins injected with the pCI E2t construct and NOB titers. Anti-E2 titers are shown as the reciprocal of the serum dilution equivalent to 3× the OD of sera from each tamarin obtained before the primary injection. Open symbols correspond to i.e.-immunized animals, and solid symbols represent i.m.-injected animals. Booster injections are indicated by vertical arrows. NOB titers, expressed as the reciprocal of serum dilution which gave 50% neutralization of binding in the NOB assay, are presented in tabular form. p.i., post-primary injection.

Surprisingly and in contrast to observations made with mice for which all previously tested plasmids failed to induce antibodies with NOB activity (data not shown), three of the four i.e.-injected tamarins developed NOB antibodies (Fig. 4). Sera with NOB activity corresponded to sera displaying the highest anti-E2 antibody titers. This was particularly unexpected in view of the fact that sera from mice injected with the same plasmid induced anti-E2 antibody titers that were always at least 1 log higher than those detected in tamarins (Fig. 2A). This finding suggests that the host specificity in addition to the form of the expressed antigen is a critical component in the induction of antibodies with biological relevance.

DISCUSSION

Components of viral envelopes include, apart from cellular proteins, specific viral antigens, mainly glycoproteins, that play a critical role in the induction of neutralizing antibodies. Interaction between such proteins has been shown to govern the accessibility of epitopes to the immune system and, therefore, the quality of the immune response mounted by an infected or immunized host (11, 31). By using eight different plasmids engineered to allow or prevent HCV E1 and E2 envelope glycoproteins to form heterodimers proposed as the prebudding form of the HCV envelope complex, together with different immunization routes, we demonstrated here that only plasmids expressing truncated forms of the antigens that have lost the ability to assemble into native noncovalent complexes are capable of inducing detectable antibodies. For the E2 protein, all plasmids encoding a truncated form of the antigen (with one exception, pCI E1E2t) were associated with a considerable benefit in seroconversion rates and antibody titers induced. The fact that the pCI E1E2t construct failed to induce a significant humoral response may be simply related to the very low level of antigen expression observed compared with those of other plasmids (Fig. 1B).

There are different hypotheses that could account for the fact that native complexed antigens fail to induce a significant antibody response. Previous reports have shown that secreted DNA-expressed antigens versus nonsecreted ones seem to induce optimum immune responses (3, 20). Indeed, in our study, the secretion of full-length E1 and E2 proteins, complexed or not, may be dramatically impaired or prevented. Thus, the proteins may remain trapped within the endoplasmic reticulum as we observed here. In vitro experiments, with recombinant viruses expressing HCV antigens or a stable cell line expressing the full-length HCV polyprotein (25–27, 36) suggest that these proteins are not secreted. The lack of secretion of the HCV glycoproteins, therefore, may have resulted in the lack of presentation of antigenic epitopes to the immune system. Many reports have nonetheless showed that, even when using recombinant viruses expressing E1 and/or E2 proteins devoid of their transmembrane domains, such as those used in our study, secretion of antigens is observed at very low levels only (26). Thus, secretion of the HCV E1 and E2 may not play the major role in the induction of the humoral response that we observed here. One other explanation is that the formation of E1-E2 complexes from native full-length proteins would result in the masking of important determinants, as described for the gp120 and gp41 subunits of the human immunodeficiency virus env protein (11), and in particular those associated with NOB activity. Because the NOB determinants remain to be mapped, this hypothesis is presently difficult to evaluate. Finally, when antigens are not or very poorly secreted, induction of antibodies may be due to the lysis of expressing cells by the specifically induced cellular immune response via the action of cytotoxic T lymphocytes (7). Although we did not analyze the induction of cytotoxic T lymphocytes in our study, the analysis of in vitro splenic cytokine production showed similar levels of IFN-γ and Il-2 secretion for all plasmids. These data are suggestive of the induction in all cases of a Th1-like response which does not appear to be enhanced by any particular plasmid. Thus, one likely hypothesis would indeed be the masking of important determinants as being the principal limiting factor in the induction of antibodies when interactive full-length E1 and E2 proteins are expressed.

Our study reveals the very low immunogenic potential of E1 when directly injected as DNA, and this observation is concordant with data recently reported by Lee et al. from the rat model (23). In chronically infected chimpanzees, anti-E1 antibodies have been efficiently raised by immunization with recombinant E1 protein and were associated with clearance of viral antigen from the liver (23b). This recent observation indicates, although yet to be confirmed, a possible role of anti-E1 antibodies in the control of liver inflamation and thus accentuates the necessity to optimize E1-expressing plasmids.

An important result achieved with constructs expressing a truncated form of E2 is the induction in mice of antibodies displaying reactivity against HVR-1 (Fig. 3), which contains an important neutralization domain (37, 44). Those were exclusively associated with the use of an i.e.-based injection route and were independent of the overall anti-E2 antibody titers induced. It has recently been proposed that the avidity of antibodies generated after the direct injection of recombinant plasmids was dependent on the immunization route (4). Although further studies concerning the avidity of the anti-HVR-1 antibodies are necessary, our results suggest that the i.e. route could induce antibodies with higher avidity than the i.m. route. Data obtained from a peptide-based scanning analysis indicated that plasmids expressing either truncated or full-length E2 antigens induce antibodies mainly directed at the C terminus of HVR-1. Although some reports suggest that antibodies directed at the C terminus alone might not be sufficient for clearance of virus because chronically infected patients contain these antibodies (35, 44), another study shows that serum reacting to the C-terminal 13 aa of HVR-1 (aa positions 398 to 410) prevented isolate-specific infection with HCV in cell culture (37). Further investigation of the significance of anti-HVR-1 antibodies for elimination of HCV is needed to determine which type of antibody would be important to induce.

The most dramatic illustration of the benefit achieved by using a truncated HCV E2-expressing plasmid was that it was possible to induce in tamarins NOB antibodies, which have been linked to the prevention of infection in chimpanzees and the control of chronic infection in humans (21, 32). Although the induced NOB titers were low, clinical resolution of hepatitis C was observed in some infected patients developing similar titers (21). In the tamarins, induction of such antibodies was dependent on the overall anti-E2 antibody titers, themselves linked to the injection route (because antibody titers were constantly found to be higher when the i.e. route was used). DNA vaccination has been demonstrated to be efficacious in a number of preclinical animal models (mice and rabbits), but data from primates have been more limited. The present study underlines the difficulty in extrapolating results observed in one species to another for a given DNA-expressed antigen. It also points perhaps to the most critical role played by the host cell’s machinery or the host-specific immune system compared with that played by the nature of the plasmid itself or the injection route for the induction of biologically relevant antibodies.

Genetic immunization has been shown to be a very powerful inducer of immune responses when single or noninteractive antigens are expressed. We demonstrate here its limitation when plasmids that express interactive intracellular antigens such as HCV envelope proteins are used. As shown in our study, induction of at least one kind of neutralizing antibody may nonetheless be efficiently generated by using a plasmid engineered to express a nonnative antigen.